05-28, 12:10–12:20 (CET), Seed
AI in the Web3 space is experiencing a boom, primarily due to the rise of LLMs. However, a critical application of machine learning, which has long been utilized in centralized payment solutions, is fraud prevention. Each time we send money or pay using a credit card, state-of-the-art algorithms decide if the transaction should go through. But not on the chain. Let's change that.
I'll talk about AI in the Web3 space and propose a new direction that is, in my opinion, worth doing -- fraud prevention. Currently, AI is mostly synonymous with LLM-based agents who are completing prompted tasks. I'll argue that machine learning includes many disciplines besides LLMs and that we should be focusing on these as well. One of them is fraud prevention, which is widely used in the standard Web2 space. Banks, big marketplaces, games, and many others all include ML-based fraud prevention systems, which decide if the given action should be taken or not. There is no reason why users of blockchain shouldn't be protected and warned about potentially malicious actions, but at the same time keep their self-custodial wallets. I'm part of Gnosis AI and currently experimenting with the project codenamed "Safe Guard Agent", which is doing exactly that.
Peter Jung, an ML/LLM Engineer at Gnosis and PhD student, specializes in blending machine learning with Web3. As the maintainer of the Prediction Market Agent Tooling library, his work focuses mainly on accurate and profitable event prediction on the chain, but also various other use-cases with potential to help blockchain users. Peter's expertise is also showcased on his web page, www.jung.ninja.